Use fences for the APPLESYNC codepath, and implement event queries
on top of fences.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Fixes a regression introduced by commit
be20ddc38b.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
In order to avoid recording vertices generated by other operations, e.g.
blits.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Dropping WINED3D_MAP_DISCARD too early means we go through the system memory
path unnecessarily. Worse, as mentioned in an earlier comment, Darkstar One
depends on the entire buffer being uploaded in case of WINED3D_MAP_DISCARD
maps, and dropping the flag too early prevents that as well. This fixes a
regression introduced by commit 15d53761a5.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Usually when doing a WINED3D_MAP_DISCARD map on a buffer with a buffer object
WINED3D_LOCATION_BUFFER will already be current. However, if the previous map
on that buffer loaded WINED3D_LOCATION_SYSMEM, that's not necessarily true.
This also brings the code further in line with the corresponding code for
textures.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Particularly for applications that use dynamic buffers without
WINED3D_MAP_DISCARD/WINED3D_MAP_NOOVERWRITE.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Mostly to avoid making things worse for applications that map dynamic buffers
without WINED3D_MAP_DISCARD/WINED3D_MAP_NOOVERWRITE.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Instead of trying to replicate the wined3d_buffer_map() logic in
wined3d_buffer_unmap().
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
It might seem like a good idea to try to use glMapBufferRange() in
buffer_direct_upload(), but it really isn't. The interesting cases for
glMapBufferRange() are GL_MAP_INVALIDATE_BUFFER_BIT or
GL_MAP_UNSYNCHRONIZED_BIT, but the only way for those to be set here
would be for the buffer to be mapped with WINED3D_MAP_DISCARD or
WINED3D_MAP_NOOVERWRITE before the buffer object was created. In that
case the GPU is obviously not using the buffer object, so
glMapBufferRange() doesn't help. On the other hand, if the buffer is
currently in use by the application and neither WINED3D_MAP_DISCARD nor
WINED3D_MAP_NOOVERWRITE is set, glMapBufferRange() would need to wait
for the GPU to finish using the buffer object, while glBufferSubData()
may decide not to.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Fixes bc019fcbbd. We cannot use
unsynchronized updates when the application haven't called map at all.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Fixes bc019fcbbd. Previously, we had the
NOSYNC flag instead of the SYNC flag. When the NOSYNC flag was set the
buffer was mapped with GL_MAP_UNSYNCHRONIZED_BIT. Now, when the SYNC
flag is not set the buffer is mapped with GL_MAP_UNSYNCHRONIZED_BIT.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Fixes a regression introduced by commit
e3a8cd20ce.
Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Józef Kucia